High frequency (HF) electrocardiography analyzing the entire QRS complex in the frequency range of 150 to 250 Hz may prove useful in the detection of coronary artery disease, yet the long-term stability of these waveforms has not been fully characterized. We therefore prospectively investigated the reproducibility of the root mean squared (RMS) voltage, kurtosis, and the presence versus absence of reduced amplitude zones (RAzs) in signal averaged 12-lead HF QRS recordings acquired in the supine position one month apart in 16 subjects and one year apart in 27 subjects. Reproducibility of RMS voltage and kurtosis was excellent over these time intervals in the limb leads, and acceptable in the precordial leads using both the V-lead and CR-lead derivations. The relative error of RMS voltage was 12% month-to-month and 16% year-to-year in the serial recordings when averaged over all 12 leads. RAzs were also reproducible at a rate of up to 87% and 8 1 %, respectively, for the month-to-month and year-to-year recordings. We conclude that 12-lead HF QRS electrocardiograms are sufficiently reproducible for clinical use.